Abstract
Fuzzy-connective-based aggregation networks are capable of aggregating information in a hierarchical manner. It simulates the human decision-making process and taking the compensation into consideration. The result can be interpreted as a set of rules that capture an abstract model of the problem. Identifying the relative importance of criteria can also help detect redundant features that do not contribute to the decision-making process. Conventional gradient-based learning approach tends to generate local solutions, and requires the aggregation function to be continuous and differentiable. In order to enhance the effectiveness and applicability, this study proposed GA-based and PSO-based learning approaches to determine the weights and parameters in fuzzy-connective-based aggregation networks. The effectiveness of our proposed methods are demonstrated using seven datasets with different number of criteria and two practical cases with regard to investment environment evaluation and location selection. Statistical analysis of the experimental results confirms that the proposed approaches outperform the conventional method, having better generalization ability and generating more accurate and reliable estimates. The proposed approach is well suited to a broad range of fuzzy aggregation connectives, which further expands its applicability.